Opened 2 months ago
Closed 2 months ago
#3232 closed task (fixed)
System upgrade matrix container to bullseye
Reported by: | robe | Owned by: | robe |
---|---|---|---|
Priority: | normal | Milestone: | Sysadmin Contract 2024-I |
Component: | SysAdmin/Matrix | Keywords: | |
Cc: |
Description
matrix container is currently at buster (debian 10), needs to be upgraded and we need to do in stages, so next upgrade will be bullseye.
I'm going to start the process now, and expect matrix to be up and down why I make this change.
Change History (3)
comment:1 by , 2 months ago
comment:2 by , 2 months ago
On further inspection I guess it's because heisenbridge depends on python3.7 which is removed by autoremove since debian bullseye ships with 3.9.
Looks like heisenberg is installed via venv via ansible? So we need to rebuild that venv.
I'm hesitant to mess with that setup unless strk is around to catch me if I fall. Looks like it uses an ansible built-in heisenbridge module, which I would have to pray knows to pull in 3.9, which I assume it would and then we would need to possibly repatch the identd.py script again for 3.9 (or not).
I'll open a separate ticket for this and will finish the autoremove and fix failed service.
comment:3 by , 2 months ago
Resolution: | → fixed |
---|---|
Status: | assigned → closed |
Heisenbridge is ok now. A problem I noticed on the machine was a bogus entry in /etc/source.list - I've fixed that (the bullseye-security url was wrong, the bullseye-updates was missing). I've upgraded again after fixing the entries so now we're at Synapse 1.112.0
I've run the autoremove and Heisenbridge is still up.
I guess we can close this ticket, feel free to reopen if anything to do is left.
I upgraded to debian-11 and also the PostgreSQL from 11 to 13.
However if I try to run apt auto-remove, then heisenberg-bridge fails to start.
If I try to run heisenberg-bridge manually after auto-remove, I gets this:
So I assume it must be pointing at an obsolete python or hard-wired to look for stuff.
To fix I reverted back to before I ran:
There is still one other service that either way doesn't start, but I think i've seen that before and is harmless, I have to look up how I fixed that:
There is also issue with nginx key being expired. My usual fix doesn't seem to work so I'll revisit later